A constrained parameter evolutionary learning algorithm for Bayesian network under incomplete and small data

• Published in: Chinese Control Conference pp. 3044 - 3051
• Main Authors: Yao You, Jie Li, Ning Xu
• Format: Conference Proceeding
• Language: English
• Published: Technical Committee on Control Theory, CAA 01.07.2017
• Subjects: Algorithm design and analysis; Bayes methods; Bayesian network; CPEL algorithm; Data models; evolutionary strategy; Heuristic algorithms; Inference algorithms; Maximum likelihood estimation; parameters learning; qualitative knowledge; Training data
• ISSN: 1934-1768
• DOI: 10.23919/ChiCC.2017.8027825

Lack of relevant data is a major challenge for Bayesian network (BN) parameters learning. For the issue, this paper proposes a constrained parameter evolutionary learning algorithm (CPEL) which is based on the qualitative knowledge and evolutionary strategy. In detail, firstly qualitative knowledge is employed into BN parameters learning process to reduce the parameter search space where two types of qualitative knowledge with experts' confidence are presented; and then evolutionary strategy is introduced into the process to avoid the problem that classical learning technique falls into local optimum easily in which the special encoding for the BN parameters is presented and some evolutionary strategies are discussed. So combining their advantages will have an important significance for BN parameters learning under incomplete and small data. Comparative experiments show that the CPEL algorithm is better than classical EM algorithm in accuracy and timeliness performance, which verify the feasibility and superiority of the algorithm proposed. Additionally, the CPEL algorithm has been applied to UAV threat assessment under complex dynamic environment.